2-D Niblett-Bostick magnetotelluric inversion - MTNet
2-D Niblett-Bostick magnetotelluric inversion - MTNet
2-D Niblett-Bostick magnetotelluric inversion - MTNet
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
J. RODRÍGUEZ et al.<br />
only the NACP anomaly. The TOBE anomaly is somehow<br />
diluted towards the right side of the model. The results are<br />
somewhat similar when only series data are inverted, as<br />
can be observed in Fig. 8B. The main difference is that in<br />
this last case, the NACP anomaly decreases to about half<br />
the size because of the removal of the parallel data which,<br />
as mentioned above, tends to emphasize conductors. This<br />
effect is explicitly shown in Fig. 8C, where only parallel<br />
data were used in the <strong>inversion</strong>. Both conductive anomalies<br />
are now clearly manifested, in reasonable agreement with<br />
the model shown in Fig. 7, which does not depend on the<br />
approximation. Notice also that, of the three models of Fig.<br />
8, the one for the parallel data has the least rms misfit. Still,<br />
a level of 52% misfit might seem too large. A corresponding<br />
model with a 22% misfit is shown in Fig. 9A. This was<br />
obtained simply by lowering the trade-off parameter from<br />
1 to ½. As expected, the higher-resolution model now<br />
includes smaller scale anomalies, but the same two main<br />
conductors still dominate the picture. Thus, summarizing,<br />
when in search of conductors as in most MT surveys, we<br />
should feed the approximation with parallel data, which<br />
emphasize conductors.<br />
To test the robustness of the algorithm we perturbed<br />
the data by multiplying each sounding curve by a random<br />
factor of 60% to simulate the static shift effect produced by<br />
small local variations in conductivity around the sounding<br />
sites. The results are shown in Fig. 9B. The reference<br />
FIGURE 9 models obtained by the RHaNN algorithm when random noise<br />
was added to the individual sounding curves of the CopRoD2 standard<br />
data set to simulate static shift effects. The models were obtained<br />
using data from the parallel mode. a) 0% of random noise added to<br />
data. B) 60% of random noise added to data.<br />
Geologica Acta, 8(1), 15-30 (2010)<br />
DOI: 10.1344/105.000001513<br />
FIGURE 10<br />
2-D <strong>Niblett</strong>-<strong>Bostick</strong><br />
zoom view of the first 5 km for the models shown in Figure 9.<br />
model is the same high-resolution model described in the<br />
previous paragraph, it is shown in Fig. 9A and has a misfit<br />
of 22%. When the soundings were perturbed by 60% the<br />
best fit achieved was 40%, much better than an expected<br />
FIGURE 11<br />
same as for Figure 9 but using a wider window.<br />
24